While many approaches exist for the automated segmentation of retinal vessels in fundus photographs, limited work has focused on the problem of separating the arterial from the venous trees. The few approaches that do exist for separating arteries from veins can be local and/or greedy in nature, making them susceptible to error or limiting their applicability to only the very largest vessels. Thus, there is a need for more sophisticated methods and systems for image analysis for separating arteries from veins.